Figure 1, layout of CMOS design.

To Corrode or Not To Corrode

by Andrzej Rucinski

There are two principal types of packages in which microelectronic devices are encased: plastic and ceramic. Ceramic packages display superb characteristics, offer good heat conductivity and environmental protection, but are expensive. Plastic packages, on the other hand, are inexpensive, but prone to several failures. One of them is the problem of corrosion caused by the lack of the hermeticity in plastic packages. Moisture present either during the encapsulation process or introduced during the life of the device penetrates the mold compound and causes corrosion. This constitutes a part of the so-called assembly level reliability (ARL) issue. ALR is a concern of many semiconductor manufacturers, including Fairchild Semiconductor. Fairchild has a special reliability facility, directed by Mark Rioux, aimed at reliability improvement of their products. Fairchild Semiconductor (http://www.fairchildsemi.com/) has approached the Department to assist them in looking into the problem of corrosion in semiconductor packages. A research team has been formed; it was comprised of two graduate students from the Turkish Navy, Osman Oruc and Cihangir Yuksel, working under the supervision of Dr. Andrzej Rucinski, UNH and Mark Rioux at Fairchild Semiconductor. The team has taken an approach that includes the development of a semiconductor device susceptible to corrosion. The corrosion was induced using a special testing method. Typically, coping with failing systems and devices is common, and nobody complains about the lack of failures. However, designing structures that are supposed to fail on demand is a difficult task. For example, a typical fault model in discrete logic is the stuck-at-fault model. It assumes that faults occur only on leads of gates and they are either stuck at zero or stuck at one. In the first case, a logic signal is stuck at zero regardless of the value of a proper logic signal. The second case assumes that the logic signal is stuck at one. The injection of faults is difficult and is accomplished using a pseudo-random number generator. The design of an "easy to corrode circuit" was not a simple task either. The process has taken several steps:

1. A research study of existing approaches in corrosion device design was conducted. The most promising example was identified from the Sandia National Laboratory (http://www.sandia.gov/). Their designs are based upon several principles: (1) a triple track structure is voltage biased to accelerate a corrosion process by introducing an electrochemical reaction; (2) a part of a device is unpassivated to expose the metallization layers to corrosion, whereas another part is still passivated for reference purposes; and (3) electronic heaters are provided to create an environment more susceptible to corrosion. The devices have been produced in several versions and in large quantity. Typically, the testing time has been no longer than 300 hours.

2. The UNH research team has designed similar devices that have been fabricated using 2 micron double metal CMOS technology provided by MOSIS (http://www.mosis.org). The number of samples was limited to 24 devices. All the samples were packaged using ceramic packages. The layout of the design is shown in Figure 1. The triple track structures are implemented in the metal 2 layer.

3. After initial functional tests of the fabricated devices were performed at UNH, half of the circuits were unpassivated. The test devices were then contaminated at the Fairchild South Portland facility and subjected to the HAST testing procedures under different testing regimes such as temperature, humidity, and biasing voltage.

4. The testing cycle took several months to complete because of logistic difficulties. The results have shown the presence of corrosion in unpassivated circuits occurring after several hours of testing as shown in Figure 2.

As a result of our cooperation with Fairchild, the company is going to incorporate the results into their own technology for reliability improvement. In addition, we have produced two master theses and launched a new research program supported by the US Air Force in the area of corrosion monitoring.

Figure 2, corrosion in unpassivated circuits.

The author would like to thank and acknowledge Fairchild Semiconductor's contribution, through Mark Rioux, in enhancing the quality of our program by exposing our students to real life problems.